ACI MATERIALS JOURNAL
Title no. 96-M70
Development of Statistical Quality Assurance Criterion for
Concrete Using Ultrasonic Pulse Velocity Method
by K. K. Phoon, T. H. Wee, and C. S. Loi
The ultrasonic pulse velocity (UPV) test has a strong potential to be
developed into a very useful and relatively inexpensive in place test
for assuring the quality of concrete placed in a structure. The main
problem in realizing this potential is that the relationship between
compressive strength and UPV is uncertain and concrete is an
inherently variable material. This paper presents a probabilistic
framework to incorporate these uncertainties rationally and
systematically. In doing so, a reliable and consistent statistical
quality assurance criterion based on UPV measurements can be
developed. The proposed method of analyzing UPV data allows the
engineer to make quantifiable conclusions regarding the quality of
concrete in question. In addition, the proposed method also allows
the engineer to strike an equitable balance between acceptable risk
and cost of remedial measures.
Keywords: compressive strength; quality assurance; ultrasonic pulse
velocity tests.
INTRODUCTION
Besides satisfying the strength requirement based on the
cube or cylinder test, the concrete placed in a structure must
be of uniform quality, and free of voids and discontinuities,
especially honeycombing. Lack of sufficient attention to
handling and placing of concrete can result in poor quality
concrete in the structure, even if good quality ready-mixed
concrete is used. When a dispute arises regarding the quality
of concrete in a structural member, tests are carried out to
affirm the strength of the concrete in question or to investigate
areas of poor consolidation and voids. Currently, coring for
samples to undergo compressive load testing is widely
adopted in the construction industry. However, it is costly and
time-consuming to carry out coring. In addition, there is a
practical limit to how many samples can be taken from a struc-
tural element without compromising its integrity. Because of
the small sample size, it is generally quite difficult to draw
reliable or statistically meaningful conclusions. There is also
the problem of having to repair the damage done to the
structure due to coring, which involves additional costs.
There is an obvious need for a reliable nondestructive test
method to complement or replace the existing destructive
means of verifying the strength of in place concrete. The
ultrasonic pulse velocity (UPV) test has this potential because
it is entirely nondestructive in nature, simple to operate, and
relatively inexpensive. However, the UPV test is currently
not widely used as a primary test for concrete quality. It is
used, in most instances, as a supplementary indicator of the
quality of concrete in a structure. This is because the relation-
ship between the compressive strength of concrete and UPV
is not simple.
Over the years, considerable research1-9 has been under-
taken to study this relationship. Sturrup et al.2 observed a
relationship between UPV and the logarithm of the compressive
568
TECHNICAL PAPER
strength. Ben-Zeitun3 and Price and Hynes,4 on the other
hand, proposed a linear relationship between UPV and
compressive strength, while Poh5 suggested that the relation-
ship is exponential. Regardless of the functional form, it is
widely recognized that the relationship is not unique, but is
affected by numerous factors such as the properties and propor-
tion of the constituent materials,2 age of concrete,3 presence of
microcracks,4 moisture content,6 and stresses in the concrete
specimens.7 In addition to the variability in the relationship,
it is also recognized that factors such as surface condition,
temperature of concrete, path length, and shape and size of
specimen can introduce extraneous variability to UPV
measurements if they are not properly controlled.8 The pres-
ence of reinforcements also affects UPV measurements.10 In
addition, UPV measurements taken from different speci-
mens are inherently variable because concrete is a heteroge-
neous material. The overall velocity of the ultrasonic pulse
passing through aggregates, hydration products, and void
space may be slightly different depending on their propor-
tion along the path of wave propagation. The inherent vari-
ability of UPV measurements generally cannot be removed,
although it can be reduced to some extent by proper mixing
prior to casting.
In view of the prevously mentioned uncertainties and the
associated difficulties in interpretation, the application of the
UPV test to quality assurance in practice is currently
limited.10 None of these studies considered uncertainties
systematically in the development of a correlation between
compressive strength and UPV, nor did they demonstrate
how the relationship could be reliably and consistently
applied to quality assurance. Engineers are therefore uncom-
fortable with the use of UPV as a primary method of
assessing concrete quality, which prevents the advantages of
UPV from being exploited fully. If a reliable quality assur-
ance criterion based on UPV is available, a larger part of the
structure can be tested quickly and cheaply to supplement
the limited test results obtained from destructive testing.
This will unquestionably become a valuable in place method
that can contribute significantly to the overall quality assur-
ance programs in large projects.
RESEARCH SIGNIFICANCE
The aim of this study is to make UPV the primary method
of assessing the quality of in place concrete while coring,
with its inherent disadvantages, would be relegated to the
role of calibrating the UPV with compressive strength and
ACI Materials Journal, V. 96, No. 5, September-October 1999.
Received May 20, 1998, and reviewed under Institute publication policies. Copyright
1999, American Concrete Institute. All rights reserved, including the making of copies
unless permission is obtained from the copyright proprietors. Pertinent discussion including
authors closure, if any, will be published in the July-August 2000 ACI Materials Journal if
the discussion is received by April 1, 2000.
ACI Materials Journal/September-October 1999
 Table 1Summary of ready-mixed concrete
K. K. Phoon is an assistant professor in the Department of Civil Engineering at the
National University of Singapore. He is one of the Principal Investigators in the mixture proportioning
High-Performance Concrete Research Laboratory. He received his PhD from Cornell
University in 1995. His research interests include application of probability and math-
Grade of concrete, MPa
ematical statistics to structural and geotechnical engineering. Concrete composition 35 55 75

T. H. Wee is an associate professor in the Department of Civil Engineering and Head

Cement, kg/m3 390 470 520
of the High-Performance Concrete Research Laboratory at the National University of Water, kg/m3 175 169 160
Singapore.
3
Fine aggregate, kg/m 764 698 620
C. S. Loi is an engineer in the Civil and Structural Engineering Division at the Public Coarse aggregate, kg/m3 1010 1020 1050
Works Department. He received his BEng from the National University of Singapore.
Water-cement ratio, % 45 36 31
Percentage of fine aggregate, % 43.1 40.6 37.1
confirming some of the locations of honeycombing and poor
concrete strength identified by UPV tests. The ability to test
many locations using UPV and make a rational decision Table 2Second-moment statistics of ultrasonic
based on a trade-off between properly quantified risk and pulse velocity (UPV) and compressive strength of
cost is the key advantage to adopting a statistical quality concrete
assurance approach. In this paper, a probabilistic framework Concrete Standard Coefficient of
that incorporates uncertainties rationally and systematically Variable grade, MPa Mean deviation variation, %
will be developed to illustrate how a consistent statistical 35 4.448 km/s 0.052 km/s 1.17
quality assurance criterion can be derived based on UPV UPV 55 4.591 km/s 0.054 km/s 1.18
measurements. The framework is general and can be applied 75 4.662 km/s 0.052 km/s 1.12
to any problem where indirect measurements (e.g., nondestruc- 35 41.75 MPa 1.74 MPa 4.17
tive test results) are used to diagnose some conditions in the Compressive 55 63.11 MPa 2.93 MPa 4.64
concrete medium (e.g., weak strength, honeycombing) that are strength
75 85.17 MPa 4.16 MPa 4.88
not readily amenable to direct examination. It will be shown
that this probabilistic approach is a natural extension of the
characteristic strength concept that is already widely used in the introduction of extraneous random errors. For example, all
practice. The proposed method of analyzing UPV data allows the cube specimens were water cured to minimize the vari-
the engineer to make quantifiable conclusions regarding the ability in the moisture content, which can seriously affect the
quality of concrete in question. In addition, the proposed UPV readings.3,8 Other controls included the constant calibra-
method also allows the engineer to strike an equitable balance tion of the UPV meter, especially at the start and end of each
between acceptable risk and cost of remedial measures. session when UPV measurements were taken.
The ultrasonic pulse passing through the concrete was gener-
TEST PROGRAM ated by a transducer serving as a transmitter held close to the
The test program was primarily designed to obtain: a) an surface of the concrete specimen. The onset of the pulse on the
accurate probabilistic characterization of the UPV and the opposite face of the specimen was detected by another trans-
compressive strength of concrete; b) a high-quality statistical ducer that served as a receiver. This direct transmission method
correlation between UPV and compressive strength at satu- was adopted in this study because the maximum energy of the
rated surface dry condition; and c) an accurate probabilistic pulse was directed at the receiver and the noise to signal ratio
characterization of the prediction errors inherent in the corre- could be minimized. Repeated measurements were taken at
lation between UPV and compressive strength. To obtain a each transducer arrangement so that the consistency and accu-
realistic assessment of concrete variability under field condi- racy of the UPV measurements were assured. The transmitter
tions, the concrete used in this study was site-bound concrete and receiver locations were also carefully marked so that
supplied by a local ready-mixed concrete supplier in Singa- repeated measurements could be taken from the same place.
pore. Three batches of concrete in grades of 35, 55, and 75 To ensure good acoustic contact, all cube specimens were
were tested. Table 1 summarizes the mixture proportion used cast using smooth steel molds so as to obtain level surfaces.
in the three mixtures. The aggregates used for the concrete Acoustic contact was further improved with the use of
mixtures were river sand and crush granite, with maximum acoustic grease. The layer of grease between the transducer
aggregate sizes of 5 and 20 mm, fineness moduli of 3.1 and and concrete surface was kept as thin as possible so that the
6.5, and specific gravities of 2.60 and 2.65, respectively. readings taken would not be too adversely affected by the
Superplasticizer was used to control the slump values of thickness of the grease. Furthermore, constant pressure was
these concrete mixtures within the range of 125 25 mm. applied on the transducers during the course of the UPV
For each batch of concrete, 50 150 x 150 x 150 mm cube measurements. By taking these precautions, extraneous vari-
specimens were cast for statistical study. All the specimens abilities introduced by poor procedural and operator controls
were subjected to nondestructive UPV tests followed by were thus reduced.
destructive compressive strength tests on the 28th day. The
equipment used in this study to measure the UPV is the portable TEST RESULTS
ultrasonic nondestructive digital indicating tester (PUNDIT) Second-moment statistics
with transducers operating at a frequency of 150 kHz. The mean, standard deviation, coefficient of variation of
UPV, and compressive strength are shown in Table 2. The
Quality control coefficient of variation is a measure of relative dispersion
Because of the statistical nature of this study, stringent and is computed by dividing the standard deviation by the
quality control had to be maintained throughout the test mean. The relatively low coefficients of variation for UPV
program to minimize the occurrence of systematic errors and ( 1.1 to 1.2%) and compressive strength ( 4.2 to 4.9%) are

ACI Materials Journal/September-October 1999 569


Fig. 1Cumulative distribution functions of ultrasonic pulse
velocity (UPV): (a) Grade 35; (b) Grade 55; and (c) Grade 75.
indications of the stringent quality control maintained
throughout the test program. The corresponding coefficients
of variation for in place concrete are likely to be higher than
those obtained in this study. It is also worth noting that the
coefficients of variation are relatively uniform across
different grades of concrete.
Cumulative distribution functions
The cumulative distribution functions of the UPV and
compressive strength for different grades of concrete are
plotted and tested for normality using the standard
Kolmogorov-Smirnov goodness-of-fit test. Figures 1 and 2,
respectively, show that the empirical cumulative distribution
functions of UPV and compressive strength can be approxi-
mately fitted to the theoretical normal cumulative distribution
function. Results from the Kolmogorov-Smirnov test shown
in Table 3 also indicate that the observed D-statistic is less
than the critical value corresponding to the customary 5%
level of significance. The null hypothesis that the data
belong to a normal probability distribution therefore cannot
570
Fig. 2Cumulative distribution functions of compressive
strength of concrete: (a) Grade 35; (b) Grade 55; and (c)
Grade 75.
be rejected at the 5% level of significance. Both UPV and
compressive strength can thus be adequately modeled as
normal random variables with the respective means and stan-
dard deviations indicated in Table 2.
STATISTICAL QUALITY ASSURANCE CRITERION
Statistical correlation between UPV and
compressive strength
The relationship between UPV and compressive strength
is shown in Fig. 3. It is observed that the compressive
strength of the concrete specimens generally increases
linearly with UPV. Such a linear relationship was also
observed by a number of other researchers.3,4 The relation-
ship between UPV and compressive strength can be esti-
mated using the standard linear regression technique. The
best fit line for the data obtained in this study is
Y = 142.4X 587.0 (1a)
where Y = compressive strength (MPa) and X = UPV (km/s).
Because of the large number of specimens tested in this
ACI Materials Journal/September-October 1999
 Fig. 4Cumulative distribution function of prediction
errors from linear regression.

Fig. 3Statistical correlation between UPV and compressive

strength of concrete. rather than the mean strength. The characteristic strength of
concrete is defined in probabilistic terms as follows
Table 3Kolmogorov-Smirnov goodness-of-fit test
for normal distribution P(Y < yc) = (3a)
Grade of Observed Critical value Probability
Variable concrete, MPa D-statistic D5% distribution
where
Y = compressive strength;
35 0.108 0.192 Normal
yc = characteristic strength; and
UPV 55 0.104 0.192 Normal
= probability that the compressive strength of concrete
75 0.073 0.192 Normal
is less than yc .
35 0.076 0.192 Normal The characteristic strength is therefore specified such that
Compressive
strength
55 0.157 0.192 Normal the fraction of concrete specimens with compressive
75 0.181 0.192 Normal strength less than yc is acceptably small. The commonly
Note: D-statistic = maximum absolute difference between empirical and theoretical accepted value of is 5%.11 For compressive strength that is
cumulative distribution function. The null hypothesis that variable is normally distrib-
uted can not be rejected if observed D-statistic is less than critical value D5%. determined by direct destructive testing of cube or cylinder
specimens, the existing practice on quality assurance [Eq.
study (n = 150), it is also possible to study the reliability of (3)] is already based on probabilistic principles that can
using Eq. (1) to predict compressive strength from UPV rationally account for the variability in the test results.
readings. The prediction errors are summarized in the form In the case of the nondestructive UPV test, a characteristic
of a cumulative distribution function as shown in Fig. 4. The value for UPV can be defined in an analogous manner as follows
mean and standard deviation (SD) of the prediction errors are
0 and 10.46 MPa, respectively. The significant errors P(X < xc) = (4a)
computed are not surprising because concrete is a heteroge-
neous material with considerable inherent variability, and where
UPV is not affected by compressive strength alone, as X = UPV;
discussed previously. xc = characteristic UPV; and
As shown in Fig. 4, the empirical cumulative distribution = probability that the UPV of concrete is less than xc .
function of the prediction errors is very similar to the theoret- From a practical engineering point of view, the evaluation
ical normal case. The observed D-statistic from the of the characteristic value of UPV xc is, however, not very
Kolmogorov-Smirnov test is 0.089, which is smaller than the useful or meaningful unless it can be related to the character-
critical value of 0.11. The null hypothesis that the data belong istic strength yc .
to a normal probability distribution therefore cannot be rejected To derive this relationship, the previously established
at the 5% level of significance. Based on the previously cited linear relationship between UPV and compressive strength is
detailed error analysis, Eq. (1) can now be reformulated more first expressed as
realistically in probabilistic terms to account for the prediction
errors as follows Y = aX + b + (5a)

Y = 142.4X 587.0 + (2a) where

where = normal random variable with mean = 0, and stan-
dard deviation = 10.46 MPa.
Probabilistic model for quality assurance
To account for the variation in the compressive strength
taken from different cube or cylinder specimens, it is
customary to use the characteristic strength of concrete
a = 142.4 (MPa s/km);
b = 587.0 (MPa); and
= normally distributed random variable with mean
= 0 and standard deviation = 10.46 MPa.
The model parameters a and b may change depending on
factors such as the type of aggregates used in the concrete. In
practice, the actual relationship between UPV and compres-
sive strength should be calibrated using some representative

ACI Materials Journal/September-October 1999 571

core samples. Based on the prediction errors observed in this
study, it is estimated that approximately 10 samples are
needed to obtain a reasonably accurate mean prediction. It
may be possible to augment this sample size with the cube or
cylinder specimens cast on site for quality control or labora-
tory specimens constituted using the same concrete mixture
to obtain a more refined mean relationship and a better esti-
mate of the uncertainty about the mean .
It was demonstrated previously that the UPV can be
adequately modeled as a normal random variable with mean
= X and standard deviation = X. Because of the normality
of X and and the additive form of Eq. (5), it can be easily
shown that the compressive strength Y also follows a normal
distribution with mean Y and standard deviation Y given by
Y = a X + b
2 2 2 2
Y = a X +
Note that the deduced normality of the compressive strength
is also consistent with the experimental data shown in Fig. 2.
The relationship between the characteristic strength yc and
the characteristic UPV xc can now be established by
rewriting Eq. (3) as follows
P aX + b + y c =
Either Eq. (9) or (12) can be used for quality assurance. Eq.
(12) is, however, slightly more robust in the sense that it is less
affected by changes in X . For example, assume that = =
0.05, a = 142.4 MPa s/km, b = 587.0 MPa, = 10.46 MPa,
and xc = 4.422 km/s. For values of X ranging from 0.01 to
0.02, the value of yc determined from Eq. (12) varies from
33.1 to 36.6 MPa. The mean UPV (X) corresponding to the
above characteristic UPV is 4.534 km/s [evaluated using
Eq. (11) with an average X of 0.015]. For the same varia-
tion in X , the value of yc determined from Eq. (9) now
varies over a larger range from 31.3 to 38.4 MPa.
Summary of proposed approach
The proposed approach for quality assurance using UPV
can be summarized as follows:
(6a) 1. Determine the model parameters a and b in Eq. (5) by
calibrating the relationship between UPV and compressive
strength using some representative core samples. If possible,
(6b) augment this sample size with the cube or cylinder speci-
mens cast on-site for standard compressive strength tests or
laboratory specimens constituted using the same concrete
mixture. In the absence of additional information, a
first-order estimate of the uncertainty about the mean rela-
tionship is 10 MPa.
2. Measure the UPV of the concrete section in question
and summarized the results in the form of the mean X and
(7a) standard deviation X as follows

Since Y = aX + b + is a normal random variable, Eq. (7) n

1
can be readily calculated as follows X = ---
n xi (13a)

y c Y y c a X b i=1
---------------
- = --------------------------------
- = (8a)
Y a2 2 + 2 n
X
1
xi X
2
X = ------------ (13b)
n1
where () = cumulative distribution function of a standard i=1
normal variate that can be readily obtained from standard
statistical tables.12 The mean UPV X is thus related to the where
characteristic strength as n = number of UPV readings; and
xi = i-th UPV reading.
1 2
y c = a X + + a X + b
2 2
(9a) 3. Select an acceptable risk level and evaluate the char-
acteristic strength yc corresponding to the UPV measure-
ments using Eq. (9).
Finally, note that X is also normally distributed and there-
fore Eq. (4) can be evaluated as 4. Alternatively, first evaluate the characteristic UPV value xc
using Eq. (11), with generally taken as 0.05. Note that the
x c X coefficient of variation X is calculated as X/X . The character-
----------------
- = (10a) istic value for strength is then determined using Eq. (12).
X
Illustrative example
or To illustrate the previous approach, consider the common
problem of checking the quality of concrete placed in a
1
x c = X 1 + X (11a) structure to see if it can fulfill a certain design characteristic
strength. It will be most desirable if extensive checks can be
where carried out quickly and cheaply using a nondestructive test
X = X/X = coefficient of variation of X. such as the UPV test. Because the sample size is much larger,
Note that the values of X obtained in this study are very statistically meaningful conclusions can be drawn. Let us
uniform (Table 2). Therefore, in practice, it may be sufficient to assume that the mean X and standard deviation X of the
determine X only, which typically requires a smaller sample UPV measurements taken from a particular section are 4.534
size, while X is simply evaluated as X X . By combining Eq. and 0.068 km/s, respectively. The coefficient of variation X
(9) and (11), the relationship between yc and xc can be expressed of the UPV measurements is 1.5%. The characteristic UPV
as follows determined from Eq. (11) based on = 0.05 is 4.422 km/s.
If the acceptable rejection rate for compressive strength is
1 2 2 2 1 1 5% (i.e., = 0.05), the characteristic strength yc implied by a
y c = a X + + ax c 1 + X + b(12a) characteristic UPV value of 4.422 km/s can be calculated from

572 ACI Materials Journal/September-October 1999

Eq. (12) as 35.2 MPa. For illustrative purposes, the model Table 4Variation of characteristic strength*
parameters a and b obtained in this study are used in Eq. (12). Characteristic strength, MPa
The characteristic strength value thus obtained can be Acceptable
risk, % X = 1.0% X = 1.5% X = 2.0% X = 2.5% X = 3.0%
compared with the design characteristic strength to determine
1 30.0 25.5 20.0 13.9 7.4
if the quality of the concrete is acceptable. If the rejection rate
is higher, e.g., = 0.10, the value of yc becomes 40.4 MPa. The 5 38.4 35.2 31.3 27.0 22.4
computed characteristic strength is more optimistic since the 10 42.9 40.4 37.3 34.0 30.4
engineer is willing to accept more risk of the concrete not 20 48.3 46.6 44.6 42.4 40.1
achieving the characteristic strength. On the other hand, if the 30 52.2 51.2 49.9 48.5 47.1
concrete is more variable, e.g., X = 0.02, the value of yc *With risk level and inherent variability of concrete for mean ultrasonic pulse velocity =
becomes 31.3 MPa. This value is more pessimistic because 4.534 km/s.
Note: X = coefficient of variation of UPV readings.
quality control is poorer as reflected by the higher X. More
examples on the variation of characteristic strength with risk
level and inherent variability of concrete are shown in Table 4. NOTATION
It can be clearly seen that the proposed quality assurance crite- a,b = slope and intercept of linear regression equation, respectively
rion systematically and rationally accounts for the inherent D-statistic=maximum absolute difference between empirical and theoreti-
cal cumulative distribution function
variability of concrete, the reliability of the prediction model, D5% = critical D-statistic corresponding to 5% level of significance
and the risk level the engineer is willing to take in the evalua- n = number of cube specimens
tion of the characteristic strength. r = correlation coefficient from linear regression analysis
SD = standard deviation
UPV = ultrasonic pulse velocity
CONCLUSIONS X = UPV, km/s
A laboratory testing program was undertaken to evaluate xc = characteristic UPV
the probabilistic characteristics of UPV and compressive xi = ith UPV reading
yc = characteristic compressive strength
strength. It was found that the coefficients of variation for Y = compressive strength, MPa
UPV and compressive strength remained relatively uniform = rejection rate for compressive strength
for the three different grades of concrete under study (35, 55, = rejection rate for UPV
and 75 MPa). Under well-controlled laboratory conditions, X = coefficient of variation of X
= prediction error from linear regression
the coefficients of variation for UPV and compressive X = mean of X
strength were observed to vary from 1.1 to 1.2% and from Y = mean of Y
4.2 to 4.9%, respectively. These statistics are likely to be X = standard deviation of X
lower-bound values for the inherent variability of concrete. Y = standard deviation of Y
Comparison between the empirical and theoretical normal = standard deviation of
() = cumulative distribution function of standard normal variety
cumulative distribution functions showed that UPV and
compressive strength could be adequately modeled as
normal random variables. This observation was further REFERENCES
1. ACI Committee 228, In-Place Methods for Determination of
supported by the Kolmogorov-Smirnov goodness-of-fit test. Strength of Concrete, ACI Materials Journal, V. 85, No. 5, Sept.-Oct.
A probabilistic model to predict compressive strength from 1988, pp. 446-471.
UPV was then developed using linear regression analysis. The 2. Sturrup, V. R.; Vecchio, F. J.; and Caratin, H., Pulse Velocity as
Measure of Concrete Compressive Strength, In Situ Nondestructive
prediction error in the model was found to be normally distrib- Testing of Concrete, SP-82, V. M. Malhotra, ed., American Concrete Insti-
uted with a standard deviation of 10.46 MPa. It is important to tute, Farmington Hills, MI, 1984, pp. 220-227.
note that the model parameters are expected to change with 3. Ben-Zeitun, A. E., Use of Pulse Velocity to Predict Compressive
field conditions. Hence, the numerical values of the model Strength of Concrete, The International Journal of Cement Composites
and Lightweight Concrete, V. 8, No. 1, Feb. 1986, pp. 51-59.
parameters reported in this paper should not be used unless 4. Price, W. F., and Hynes, J. P., In Situ Strength Testing of High-
there are some field data to substantiate their validity. Using Strength Concrete, Magazine of Concrete Research, V. 48, No. 176., Sept.
this probabilistic model and the statistics of UPV determined 1996, pp. 189-197.
previously, it was demonstrated that a consistent statistical 5. Poh, C. S., Evaluation of Ultrasonic Pulse Velocity Method, masters
quality assurance criterion using UPV could be developed. The thesis, National University of Singapore, 1989.
6. Swamy, R. N., and Al-Hamed, A. H., Use of Pulse Velocity Measure-
proposed statistical quality assurance criterion was shown to be ments to Estimate Strength of Air-Dried Cubes and Hence In Situ Strength
a natural extension of the characteristic strength concept that is of Concrete, In Situ Nondestructive Testing of Concrete, SP-82, V. M.
already widely used in practice. In addition, the statistical quality Malhotra, ed., American Concrete Institute, Farmington Hills, Mich., 1984,
assurance criterion also rationally and systematically took into pp. 247-276.
7. Popovics, S.; Rose, J. L.; and Popovics, J. S., Behavior of Ultrasonic
account the inherent variability of concrete, the reliability of the Pulses in Concrete, Cement and Concrete Research, V. 20, No. 2, Mar.
prediction model, and the risk level the engineer is willing to take 1990, pp. 259-270.
in the evaluation of the strength of in-situ concrete and in the 8. BS 1881 Part 203:1986, Recommendations for Measurement of
investigation of areas of poor consolidation and voids. This Velocity of Ultrasonic Pulses in Concrete, British Standards Institutions,
proposed method allows the engineer to strike an equitable 1986, pp. 1-16.
9. ASTM C597-83, Standard Test Method for Pulse Velocity
balance between acceptable risk and cost of remedial measures. through Concrete, ASTM, Philadelphia, PA, 1983, 3 pp.
Development of a statistical quality assurance criterion using 10. Mehta, P. K., and Monteiro, P. J. M., ConcreteStructure, Properties
other nondestructive methods is currently underway. and Materials, 2nd Edition, Prentice-Hall, NJ, 1993, p. 348.
11. ACI Committee 214, Recommended Practice for Evaluation of
Strength Test Results of Concrete (ACI 214-77), American Concrete Insti-
ACKNOWLEDGMENTS tute, Farmington Hills, MI, Part 1, 1989, pp. 1-14.
The authors would like to express their gratitude to Goh Siew Huat, Director 12. Murdoch, J., and Barnes, J. A., Statistical Tables for Science, Engi-
and Group General Manager, and his staff from Supermix Concrete Pte., Ltd., neering, Management and Business Studies, 3rd Edition, MacMillan, Hong
for their support rendered to this study. Kong, 1986.

ACI Materials Journal/September-October 1999 573